The Trade Machine
A better trade starts with a better question: what does this do for each team's starting lineup and season? Offseason combines your league's scoring, rosters, schedule, and synced history with dated player forecasts and market values. It finds plausible offers, explains the tradeoffs, and shows where the answer depends on an assumption.
The best player and the best trade are sometimes different.
A star has real value and real negotiation appeal. But a team with two weak starting positions may improve more by exchanging that star for two dependable starters. Another team may have the depth to make the opposite move.
Offseason evaluates who actually enters the starting lineup, who loses a spot, and what each position gains or gives up. Bench points do not count as automatic weekly improvement. Bye weeks, player availability, roster cuts, and processing time can change the answer from one week to the next.
A trade needs more than a value total.
Does it help my team?
See projected starting-lineup changes by week and position. Where supported, compare playoff and championship outcomes before and after the trade.
Does it make sense for them?
Compare their roster needs, modeled season outlook, market exchange, and observed activity. Offer fit is a heuristic unless outcome calibration is explicitly available.
Is there a better move?
Search practical packages across partners. Compare consolidation, improving multiple starters, and a conditional no-trade wire alternative when a fresh pool is observed.
League context meets dated football evidence.
The actual decision
Synced rosters, scoring rules, lineup slots, weekly results, future opponents, playoff settings, trade review rules, and observed manager history.
The changing baseline
Dated weekly raw-stat projections, player status, NFL schedules, completed production, observed opportunity, and format-matched trade-derived market values.
Refreshes are scheduled, and coverage is measured for the players and weeks being modeled. An unrelated fresh projection does not make a league's forecast complete. Missing or stale provider forecasts use a labeled scoring-aware fallback; missing data never become a fabricated professional projection.
Targets, carries, air yards, and expected fantasy points help describe evolving roles and production. These diagnostics are collected and compared with later outcomes. They do not silently overwrite a provider's forecast before the adjustment has been validated. Route participation is not claimed when the source does not supply current-season routes.
One scoring engine, throughout the analysis.
Raw football statistics flow through the league's scoring context. A reception, passing touchdown, interception, or defensive result is worth what your league says it is worth.
Past results prefer the platform's reported points when available. The forecast permits negative fantasy scores. Missing observations are distinct from real zeroes. Unsupported formats and unverified bonus scoring carry visible limits; bonus thresholds are not turned into a precise expected bonus without the information needed to model them.
Historical scoring reconstruction and future prediction are different tests. Reproducing an official past score does not establish forecast accuracy.
From player evidence to a reviewable recommendation.
Scoring and identity
Resolve players and league rules through shared, canonical readers.
Weekly forecasts
Use dated provider raw stats where covered, scoring-aware shrinkage where absent, and explicit availability states.
Starting lineups
Optimize legal lineup slots using the information available before the outcome. Account for uneven packages and necessary cuts.
Paired season scenarios
Compare the same simulated football futures before and after the trade. Recompute affected teams and reuse unchanged teams.
Market appeal
Keep format-matched market value alongside team benefit. Identify who receives the strongest market asset without treating that alone as the right answer.
Partner context
Use observed activity and model-based team needs. Sparse history stays sparse; quiet managers remain eligible partners.
Recommendation and explanation
Prioritize practical packages, verify a bounded shortlist with the shared season model, and expose conditional results and alternatives.
One frozen input and one baseline support a recommendation run. Semantic changes to forecasts, scoring, standings, rosters, schedules, or model policy invalidate reuse. A repeated fetch of unchanged content does not by itself require another simulation.
Find a useful exchange, then test the story.
Consolidate when depth permits
Package useful pieces for a stronger starter when the remaining roster can absorb the loss. An open spot has value, but an unverified future waiver claim is not credited as a certain acquisition.
Spread value across starters
Give up the strongest individual asset when two incoming starters improve the actual lineup enough. The explanation includes the weaker position as well as the positions that improve.
Match the season objective
A strong qualification outlook can support prioritizing playoff strength. A fragile qualification path can make upcoming weeks more important. Those objectives choose the weeks used in a bounded, heuristic ranking priority. They are modeled team needs, not diagnoses of panic or claims about a manager's intentions.
Compare the wire
A fresh observed player pool supports a hypothetical no-trade add comparison. Claim eligibility, budget, and processing time may remain unknown. A competing wire path can temper the recommendation instead of being hidden.
Simple exchanges lead the search, including both directions of a two-for-one. Balanced and other supported packages remain available when they have a clear roster purpose. A recommendation is a bounded search result, not a promise that every possible package was exhaustively evaluated.
What it trades for. What it does for your season.
Market value helps explain price, fairness, and negotiation appeal. Modeled season benefit measures changes in playoff and title outcomes against the same football scenarios. Projected lineup points provide a separate, week-by-week explanation.
The mature in-season analysis can lead with modeled season benefit. Early-season analysis retains its explicit 85% market and 15% season-score blend, with both components visible. Draft-only, incomplete, and unsupported cases keep a market read or an unavailable state rather than manufacturing season odds. Packages containing draft picks do not claim player-only season odds for the whole deal.
A useful model makes its conditions visible.
Injuries without invented recovery dates
Current designations inform the current week. An uncertain return creates explicit conditional-return and unavailable-for-the-remaining-weeks trade scenarios. No recovery date or return probability is guessed.
Timing without retroactive points
Configured review periods and known kickoffs determine the modeled effective week. Completed starter scores remain with their original team. Unknown review timing can prevent a season verdict.
Updated information, honest live limits
Known final-game scores can condition the current-week model. During active games, a full remaining-game forecast requires reliable progress data. If those data are unavailable, the tool withholds season odds and can retain future lineup evidence.
Simulation precision is not prediction certainty
Paired scenario frequencies describe this model's futures. Reported simulation error accounts for antithetic draw pairs; it does not measure injury, provider, or model misspecification error.
Verified recommendations distinguish modeled mutual gain, conditional benefit, tradeoffs, and inconclusive results. Reserve slots and uncertain injury returns can require another cut upon activation; those cases stay conditional when active-roster capacity cannot be verified. Unverified candidates carry a market or lineup read. Even a result that helps both teams across tested injury cases remains a forecast, not a guaranteed win-win.
Test correctness now. Earn accuracy claims over time.
Automated trust gates cover scoring, legal lineups, weekly availability, small-package fit, negative scores, trade timing, shared baselines, cache invalidation, and honest missing-data states. These establish implementation behavior, not a league-wide prediction success rate.
Immutable source revisions and frozen analysis inputs preserve what was known at a particular time. Historical evaluation requires a matching pre-outcome snapshot and censors incomplete outcomes. Today's player status cannot stand in for yesterday's information.
Offer outcomes distinguish observed acceptance, decline, execution, and unresolved disappearance. A vanished pending offer is not automatically a rejection. Acceptance calibration and forecast improvements must be evaluated on later, held-out observations before their accuracy is advertised.
A foundation for forecasts that keep learning.
The next accuracy gains depend on accumulated evidence: forward tests of weekly means and intervals, position-specific role adjustments, and genuinely labeled offer outcomes. Those data can justify more adaptive forecasts and better offer ranking as teams establish new baselines.
Reliable in-game progress, richer current-season participation, and professional projection feeds are potential additions subject to coverage, licensing, and validation. They are not represented as current capabilities or promised launch dates.
A trade is a change to two rosters, across a season.
Keep the price, the football benefit, and the uncertainty visible.
This edition describes the implemented method in October 2026. Availability depends on synced league data and source coverage. Fantasy data provided by Yahoo Fantasy. Football evidence includes nflverse; market values use FantasyCalc. Offseason is independent of these providers. Forecasts and offer-fit estimates are labeled model outputs.